Vibe coding: a 2-minute briefing
Forget low-code—the future of software development is pure vibes.
There’s a shift happening in how software is built—and it’s happening fast.
The term vibe coding was coined by AI researcher Andrej Karpathy, but it’s already taken on a life of its own. It’s an infinitely better name than “low-code” or “no-code,” which feel outdated and limiting. Vibe coding is not about replacing coders—it’s about working with AI to build software in an entirely new way.
Instead of manually writing every line, developers (and non-technical people) now act as curators and directors using AI-powered tools like Cursor, WindSurf, and Claude to generate and refine code at an unprecedented pace.
It means that someone can use these tools to make an app or a website in hours (minutes even) without having ever having coded in their life before.
Why the term 'vibe coding' is sticking
We’ve been talking about this shift for a while, but the phrase “vibe coding” captures something deeper than just AI-assisted development:
It’s intuitive, playful, and iterative – Instead of laboriously crafting each function, engineers now “feel” their way through a problem, prompting, tweaking, and regenerating code until it just works.
It removes the gatekeeping of traditional coding – Vibe coding is exactly what many of us do already: if you can’t code, but know how to ask the right questions, you can probably direct AI to get the job done.
It reflects a culture shift – Just as AI tools like Midjourney let artists iterate on an image by simply “rolling the dice,” engineers are now generating and discarding code at speeds that were unimaginable a year ago.
How people are actually using ‘vibe coding’
This is no longer experimental. It’s already changing how companies operate. Some fascinating real-world stats and stories:
95% of new Y Combinator startup founders say they use AI-generated code—and don’t see it as a weakness. Even highly technical founders are writing less and reviewing more.
A viral example: One developer built a multiplayer flight simulator game in just a few hours using AI-generated code, making $1,000 a month off a game that barely existed the day before, proving that with vibe coding, complex software can be built at lightning speed. More vibe coding innovation examples here.
In our own team: We’ve used vibe coding to rapidly generate interactive reports, dashboards, and even software prototypes. Recently, we used Claude and Cursor to build an entire reporting tool—writing a detailed plan, defining the input/output, and watching the AI do the rest. What would’ve taken weeks was working within an hour.
Gantt chart frustration turned into a surprise product: One of us was struggling to create a project plan, so we fed the brief into several AI tools. It didn’t give us a perfect Gantt chart—but it did generate an unexpected functional piece of software we could use for tracking revenue. Accidental innovation? That’s vibe coding.
Why it matters
This isn’t just about speeding up coding. It’s about redefining the role of a software engineer.
The rise of the "product engineer" – Code is no longer the bottleneck; taste, intuition, and problem-solving are. The best engineers are becoming product strategists first.
A massive shift in hiring – The ability to debug and architect systems is now more valuable than pure coding speed.
Zero-to-one innovation is accelerating – Building an MVP has gone from months to days, or even hours.
What has this got to do with AI?
Everything. Vibe coding is only possible because of AI models that can now write entire applications in response to a well-structured prompt.
AI can generate code, but it’s not great at finishing it (yet) – Human engineers are still needed to fix, refine, and architect.
LLMs are shifting from helpers to co-creators – Tools like DeepSeek, Claude 3.5, and GPT-4 are enabling developers to work at 100x speed, not just 10x.
The real skill is knowing how to guide AI – The difference between a mediocre and a great AI-powered engineer is not technical knowledge, but the ability to write detailed briefs and refine AI-generated results.
What does this mean for the future?
More people will "code" without traditional coding skills – If you can write a good prompt, you can build software.
Companies will need to rethink hiring – Do you need coders, or do you need system thinkers?
Startups will move faster than ever – The barrier to launching a new product is dropping to almost zero.
The best engineers will be those who can ‘see the matrix’ – Understanding how to work with AI, when to step in, and how to refine outputs will be the key skillset of the future.